Choosing a video model: length, sound and price
Start from what the shot needs: how long it runs, whether someone speaks, its frame, a picture to start from and its price. Three steps, then six video models side by side.
1. Start from what the shot needs
Say what the shot is: how long it runs, whether someone speaks, where it will be watched. On Auto, Thisbe picks the model for the job and says why on the Director's card.
2. Or choose it yourself
To pick one, open Models in the chat. Parameters, beside it, sets the length, the frame and the size.
3. Check the price before it runs
Generate on the Director's card carries the price, and no render is charged until you press it. Change the model, the length or the size, and the price is worked out again before anything runs.
For the longest takes
Of the six video models in this guide, Seedance 2.5 and Wan 3.0 run longest, up to 30 seconds a take. Kling 3.0 Pro, Seedance 2.0 and H3 run up to 15 seconds. Veo 3.1 runs up to 8 seconds, in 4-, 6- or 8-second clips. The others take any whole second in their range.
For spoken lines
Of the six video models in this guide, Veo 3.1 speaks a line you write in quotes, lip-synced, in its own sound. Seedance 2.0, H3, Seedance 2.5 and Wan 3.0 add sound of their own too, but Thisbe lists only Veo 3.1 as speaking a line you write. Kling 3.0 Pro makes its films without sound in Thisbe.
How to write one is in the guide to spoken lines.
For vertical frames
Every one of the six video models in this guide makes vertical 9:16 clips, for a phone held upright. Seedance 2.0, H3 and Seedance 2.5 take six frames, from 21:9 to 9:16. Wan 3.0 takes five frames, from 16:9 to 9:16. Kling 3.0 Pro takes three frames: 16:9, 9:16 and 1:1. Veo 3.1 takes two frames: 16:9 and 9:16.
From a picture
Every one of the six video models in this guide can start from your picture, and every one can also end on a picture you choose. H3 takes a picture up to 2.5:1, long side to short, and Wan 3.0 up to 8:1. H3 adds 3 credits for each picture it's given.
Step by step: how to turn a picture into a video.
With references
For clips, of the six video models in this guide, Seedance 2.0 takes up to 9 references, Seedance 2.5 up to 30 and Wan 3.0 up to 10. Veo 3.1 takes up to 3, on its Lite and Fast tiers and on 8-second clips only. Kling 3.0 Pro and H3 take none.
A clip takes references or a start and end frame, never both: Thisbe refuses a job that sends both before anything is charged.
For the lowest price
Of the six video models in this guide, each at the size it opens on, the lowest first take is Veo 3.1's: a 4-second Lite take at 1080p, for 27 credits. Veo 3.1 is priced by the clip, the same at any of its lengths. The other five are priced by the second, so a longer take costs more. Priced by the second, H3 and Wan 3.0 start lowest, at 30 credits for a 5-second take.
Six video models, side by side
| Model | Lengths | Sound | Priced | First take | Credits |
|---|---|---|---|---|---|
| Kling 3.0 Pro | 3s to 15s | No | By the second | 5s at 1080p | 68 |
| Seedance 2.0 | 4s to 15s | Yes | By the second | 5s at 720p | 154 |
| Veo 3.1 | 4s, 6s or 8s | With lines | By the clip | 4s Lite at 1080p | 27 |
| H3 | 4s to 15s | Yes | By the second | 5s at 768P | 30 |
| Seedance 2.5 | 4s to 30s | Yes | By the second | 5s at 480p | 105 |
| Wan 3.0 | 2s to 30s | Yes | By the second | 5s at 480p | 30 |
What each is for
- Kling 3.0 Pro. Long takes from 3 to 15 seconds, in three frames.
- Seedance 2.0. Video with sound from one prompt, 4 to 15 seconds.
- Veo 3.1. Lite, Fast or Standard, in 4, 6 or 8 seconds.
- H3. Clips with sound, 4 to 15 seconds.
- Seedance 2.5. 480p to 1080p video with sound, 4 to 30 seconds.
- Wan 3.0. Video with sound, 2 to 30 seconds, up to 1080p.